EVALUATION OF DIFFERENT ALGORITHMS FOR COLOCATION DETERMINATION

Kolokacije so besedne zveze, ki se v besedilih pojavljajo pogosteje kot bi se po naključju. V diplomskem delu bomo spoznali njihov pomen in uporabo pri procesiranju besedil v slovenskem jeziku. Pogledali si bomo tudi korpus jos1M, ki ga bomo uporabljali kot vhod v algoritme za določanje kolokacij. Implementirali bomo dva algoritma za določanje kolokacij (frekvenčni in razpršeni). Z morfološkim filtriranjem bomo izrazili kolokacije. Na koncu bomo algoritma primerjali.Collocation defines a sequence of words or terms thatco-occure more often that would be expected by chance. We will explain a meaning of collocations and their usage in processing of Slovenian text. We also describe jos1M corpus which is used for input intoalgorithms for determining of collocations. Two different algorithms for the determination of collocations (frequency and the Mean and Variance algorithm) is implemented and compared in the practical part of the thesis